In a large enterprise organizing an information system department, the computer system is operated and managed by expert system engineers well versed in computer systems. In order to maintain the performance such as response and through-put at high level, performance diagnosis of resources composing the computer system such as central processing unit (CPU) and magnetic disk devices (called disks hereinafter) is executed as routine work during the management. If performance drop is found in any resource, the system engineers investigate the cause and properly upgrade that resource.
There are enterprises which can not employ exclusive system engineers because of limited budget. Such enterprises must consign operation and management of the computer system to another company that is specialized in the maintenance work (here after maintenance firm). In this case, therefore, the performance of resources is diagnosed regularly (or irregularly) by customer engineers dispatched from the maintenance firm.
In such operation management system, however, it is actually difficult to diagnose the performance in a thoroughly understanding state of the computer system of the specific enterprise and a customer engineer engaged in performance diagnosis may overlook the cause of performance drop of resources. Under this condition, even if the resource is upgraded, the performance does not improved and the quality of customer service may be lowered. Hitherto, therefore, there has been a keen desire for development of means and method of enhancing the quality of customer service without depending on the skill of the customer engineers engaged in the operation management tasks.
Conventionally, in operation and management of a system composed of computers, the manager judges the performance on the basis of the utility rate about the system resources (CPU, disks) that compose the system. When the utility rate exceeds a threshold, the manager judges that the performance of the system resource has lowered. In such as case, the deteriorated system resource is replaced by a new system resource of higher performance, or another system resource is added.
For example, if the case the system resource is the CPU, then the present CPU is replaced by a CPU having a higher frequency than the clock frequency of the present CPU. If the system resource is a disk, another disk is added.
Conventionally, as mentioned above, when the performance of the system resources (CPU, disks) is lowered, an attempt is made to avoid the drop of the performance by replacing the existing system resource with a system resource of a higher performance, or adding new system resources.
Generally, only when the cause of performance drop of the system resource is correctly investigated the performance of the system resource can be maintained above a specified level. The conditions that lower the performance of the CPU, for example, include high utility rate and long queue waiting for execution of a process. If these are the causes, then it is effective to install an additional CPU. On the other hand, if a high utility rate is the only cause, then it is effective to replace the existing CPU with a better CPU.
The conditions that lower the disk performance include high utility rate, long queue waiting for read/write demand and long response time. If these are the causes, it is effective to install additional disks. On the other hand, if high utility rate and response time are the only causes, then it is effective to replace the existing disk with a better disk.
However, in order to judge the cause of performance drop of system resources, the engineers are required to have a thorough knowledge of the system under investigation. For the enterprise having the performance of system resources diagnosed regularly (or irregularly) by the customer engineers dispatched from the maintenance firm, in reality, it is very difficult to diagnose the performance in a state of full understanding of the computer system of the specific enterprise.
Therefore, the customer engineer responsible for performance diagnosis may overlook the cause of performance drop of system resources. In such a case, even if the system resource is upgraded, the performance is not improved so much. Regarding the CPU, for example, supposing the queue waiting for execution of process is one of the causes that generate the drop in the performance, even if the CPU is replaced with a better CPU, the performance of the CPU is not improved so much. Similarly, in the case of the disk, even if disks are added although the utility rate and response time are the causes that generate the drop in the performance, the disk performance is not improved so much.
Thus, in the event of wrong judgement of cause of performance drop of system resources, not only the quality of the customer service is lowered, but also it is accompanied by wasteful extra steps such as redoing of diagnosis and follow-up of the user by the sales staff.
Moreover, when upgrading the system resources, the system resources are ordered manually. This may cause ordering errors due to human interaction, or may take time in the ordering process.
The present invention is achieved in the light of the aforesaid problems. It is an object of this invention to provide a system diagnosis apparatus, a system diagnosis method and a computer-readable recording medium recording system diagnosis program, capable of determining measures against drop or shortage of performance of system resources and ordering necessary system resources accurately and promptly.